Tyler Cipriani's cover letter says that --force-if-includes checks the reflog of the local branch named after the destination, whatever is actually being pushed. In the code, check_if_includes_upstream() uses ref->name where it should use ref->peer_ref->name. For git push --force-with-lease --force-if-includes origin src:main, it therefore reads the reflog of main instead of src.

Cipriani describes two failure modes. If src has integrated origin/main but the local main is out of date or missing, the push is falsely rejected. If the local main has integrated origin/main but src is unrelated or outdated, the check passes and the push can clobber the remote main. He cites two earlier false-rejection reports: Stefan Haller's in July 2023 (a local branch with a different name) and D. Ben Knoble's in May 2025 (a detached HEAD). Proof-of-concept cases are in t/t5533-push-cas.sh, and he says the new tests fail against maint and pass with the fix.

Other patches in the series address related problems. In set_ref_status_for_push(), the reflog reachability check runs before the fast-forward rules, so a valid fast-forward can be rejected when the reflog is expired. Cipriani's example is a push rejected with "remote ref updated since checkout" after git reflog expire --expire=all --all. He argues this contradicts the documentation, which says the option forces an update only if the remote-tracking tip has been integrated locally, and proposes deferring the reflog rejection.

A further patch targets the advice text. When a push is rejected for lack of a reflog to consult, git suggests git pull, which would not fix the rejection. The patch message says the user needs to specify the expected tip with --force-with-lease=<ref>:<expect> or take another route around the check.

Junio C Hamano, Patrick Steinhardt, Ben Knoble and D. Ben Knoble took part in the thread, but the excerpts do not show their positions. The series is at v1 and review is still open.